SWE-QA-Pro - Correctness: leaderboard
Metric: Correctness score (1-10) with the SWE-QA-Pro agent workflow, on 260 questions from 26 long-tail repositories with executable environments, each answer scored 1-10 by a GPT-5 judge (three runs averaged) against a Claude Code reference answer checked by human annotators; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.5 | 7.34 | #138 |
| 2 | Gemini 2.5 Pro | 7.12 | #145 |
| 3 | DeepSeek V3.2 | 6.94 | #198 |
| 4 | GPT-4.1 | 6.86 | #240 |
| 5 | Devstral Small 2 | 6.61 | #484 |
| 6 | GPT-4o | 5.59 | #333 |
| 7 | Qwen 3 32B | 4.99 | #424 |
| 8 | Qwen 3 8B | 4.52 | #667 |
| 9 | Llama 3.3 70B Instruct | 2.84 | #520 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=swe-qa-pro-correctness · How It Works · Data refreshed daily, snapshot 2026-10-11.